SEIL: Simulation-augmented Equivariant Imitation Learning
SEIL: Simulation-augmented Equivariant Imitation Learning
复制标题
DOI：
10.1109/icra48891.2023.10161252
复制
发表时间：
2022-10
期刊：
影响因子：
--
通讯作者：
Ming Jia;Dian Wang;Guanang Su;David Klee;Xu Zhu;R. Walters;Robert W. Platt
中科院分区：
文献类型：
--
作者：
Ming Jia;Dian Wang;Guanang Su;David Klee;Xu Zhu;R. Walters;Robert W. Platt
在机器人操作中，获取样本是非常昂贵的，因为它经常需要与现实世界进行交互。传统的图像级数据增强已经显示出在各种机器学习任务中提高样本效率的潜力。然而，图像级数据增强不足以使模仿学习代理在合理数量的演示中学习到良好的操作策略。我们提出了模拟增强等变模仿学习（SEIL），该方法结合了一种新的数据增强策略，即用模拟过渡补充专家轨迹，以及利用机器人操作中的O(2)对称性的等变模型。实验评估表明，我们的方法可以在10次演示中学习非平凡的操作任务，并且性能明显优于基线。
In robotic manipulation, acquiring samples is extremely expensive because it often requires interacting with the real world. Traditional image-level data augmentation has shown the potential to improve sample efficiency in various machine learning tasks. However, image-level data augmentation is insufficient for an imitation learning agent to learn good manipulation policies in a reasonable amount of demonstrations. We propose Simulation-augmented Equivariant Imitation Learning (SEIL), a method that combines a novel data augmentation strategy of supplementing expert trajectories with simulated transitions and an equivariant model that exploits the O(2) symmetry in robotic manipulation. Experimental evaluations demonstrate that our method can learn non-trivial manipulation tasks within ten demonstrations and outperform the baselines by a significant margin.